FriendCare-AAL: a robust social IoT based alert generation system for ambient assisted living

The use of advanced communication technologies such as Internet of Things (IoT) in the domain of Ambient Assisted Living (AAL) tends to promote the quality of living for elderly staying independently. However, the state of the art IoT based solutions for AAL systems have not fully expressed the importance of building social connections between smart devices. This paper attempts to study the significance of deploying socially enabled IoT systems in AAL environment by proposing a robust Social IoT based AAL system for elderly named FriendCare-AAL. In addition, it presents a schematic approach to establish a partnership among smart devices and introduces the concept of responsibility offloading between devices. The proposed system is capable of providing assistance to the elderly staying in smart home environment. In case of emergency, the system automatically generates alerts intimating about the situation to the concerned entities. To experimentally evaluate the system’s performance, a smart home AAL environment for an elderly person is simulated using human activity simulator namely ‘Home Sensor Simulator’ and person’s routine dataset is generated. Further, two machine learning models; Naive Bayes (NB) and Random Forest (RF) are employed to analyze the data in order to predict the well being of the elderly person. The performance of the two classifiers is assessed using metrics such as sensitivity, specificity, detection rate and accuracy. Experimental results revealed that RF classifier outperforms NB classifier in terms of overall accuracy, detection rate and balanced accuracy. The overall accuracy is observed to be 89.2% for RF and 83.9% for NB classifier. Furthermore, a performance comparison of the proposed model is performed with two baseline approaches. A system prototype is also developed using Node-Red simulation tool to determine the performance of the proposed system in real-world and failure-prone environments. It turns out that the system performs well in critical situations with a tolerable response time of less than 1.2 s for a high failure rate of upto 50%.